Hacker News Reader

Comment by skohan | Hacker News Reader

Intuitively it makes sense to me that more complex systems collapse faster because potentially they have more points and modes of failure.

For instance, due to advances in technology, many aspects of our current modern global society depend on the availability of Cobalt as a component in battery tech. This is a relatively rare element, with concentrated extraction.

Lots of things in modern society depend on having small, cheap, powerful batteries. Lots of systems are built upon systems which have that dependency. This creates a single point of failure which can have reverberating effects throughout the whole system.

I suppose the more complex and interconnected a system is, the more likely it is that you have many of these weak points floating around.

View on HN

Complex systems indeed have a lot of failure modes, but what you'll find in any complex system which has survived any appreciable amount of time is that it'll be very tolerant of them: one maxim of system design is a complex system will always be operating in at least one failure mode: even more complex systems will usually have multiple going on at once. Systems can and do deal with this without complete collapse all the time: complete collapse usually happens only after enough failures build up over time such that it cannot compensate.

That said, this doesn't stop someone from designing a complex system which cannot tolerate failure: it's just it'll tend to fall apart as soon as they start to put it together (and they'll either learn quickly how to make it tolerate failure or it'll never get off the ground), it won't generally run fine for a long time and then implode suddenly.

Complex systems tend to have complex control systems and feedback loops that compensate for perturbations and try and maintain homeostasis.

But when these feedback loops break, for whatever reason-- reaching the limits of the elasticity and controls, for instance-- they tend to abruptly de-compensate and fall apart.

This looks sudden: more and more slack is being taken up, until the actual compensatory mechanism is exhausted, and then smack, it's all over.

It's like when you get hypothermia, and then vasoconstriction to maintain core temperature stops... and you start shedding lots of heat through your skin, and shivering ends... then you feel hot and want to take off your clothes. Your body is doing many different kinds of things to try and fight getting cold, and once one falls apart "because it's too cold" -- the rest fall like dominoes.

yes and no. complex systems are not designed to be complex from the start. they are in fact successful simple systems that have evolved. Any complex system that is around for a while will have build in redundancy and, as a matter of fact, will run in degraded mode.

So I don't buy into the complex system collapse faster idea. I would say that if you were to look at a simple system and and a [working] complex system the simple system is going to collapse faster in case of a failure, while you may not notice failures as a complex system works around them). What the author here observes is the catastrophic collapse in the late stage of the system where something leads to the almost simultaneous collapse of multiple subsystems.

Here is one of my favourite writings that ties complex systems with failure: https://how.complexsystems.fail/

You can have a complicated system that spends its complexity "budget" on redundancy and/or protection mechanisms. This means that when A fails, B keeps things working, and when B fails, C does. And so the system just keeps going, with B failed, and F, and K, and Q and X and Z. And maybe nobody (or very few people) notice that there's all these failing subsystems adding up.

And then A fails, but hey, it's still running great!

And then C fails. And the system collapses, because A, B, and C all failed. And everybody thinks that it collapsed quickly, because nobody thinks of the collapse as starting when B failed.

TL;DR: A complex redundant system can run for a long time in a partially-failed state. If you measure only from the start of full failure, you can miss how long the collapse took.

It seems to me like this relates to how much emphasis there is on competition monopolizing the entire market and eliminating less efficient solutions. In the 1980s it seemed like every quirky solution would kind of survive, maybe being too bulky and too expensive (to fabricate) but not licensing X, Y or Z for an exorbitant price.. That greatly added complexity but meant redundancies.

In the current form, I feel like the highest efficiency solution maker (or maybe the one in second place) is usually trying to do extremely low licensing with the idea of winning the whole market. That's great in terms of efficiency and actually lowers complexity but means monoculture with exactly identical dependencies.

> Intuitively it makes sense to me that more complex systems collapse faster because potentially they have more points and modes of failure.

For me, the word “complex” complicates the assertion without adding much value.

I prefer, “Systems fail faster as the number of modes and points of failure increase.”

Reply on news.ycombinator.com